Computer vision algorithms competition. Prizes go to all

Exactly a month ago, we announced a competition to create an application that implements several algorithms of machine vision. The main prize in this competition was a trip with our team to Sri Lanka to meet either the end of the world on December 21 or the New Year (this is how lucky). But we hope for the New Year more and in any case we will try to organize a broadcast from the scene.

In general, I really do not like all sorts of contests. More precisely, I treat them with some suspicion that, as a result, you will spend time on them and will not receive anything.
')
I and my colleagues at Ivideon would not like someone from those who took part in our competition to regret the time spent on it. And here we are talking about 9 potential candidates who submitted the work.
Therefore, we made a decision that 8 people who for one reason or another did not take the first place will receive for their work a megapixel IP camera with Ivideon cloud support on board. For those who do not know, this is an ordinary IP camera, in which there is our module that allows you to directly connect it to the Ivideon cloud without additional applications and computers. We do not manufacture the cameras themselves. We provide the opportunity to embed this module to all manufacturers. Remote access to such a camera does not require an external IP address and network settings such as port-forwarding. Well, it has all the features that Ivideon provides. From recording video to our cloud, to organizing a broadcast on your website or blog . We hope very much that this camera will be a worthy compensation for the time spent on our competition. Moreover, in addition to her, the participants received at least additional experience in developing video analysis applications.



All the work sent, we conditionally divided into three groups:

1. Examples from the Internet, which, although solving partly assigned tasks, still remain examples.



Moreover, they clearly trace the absence of a single code formatting style, which indicates a quick copy-paste from various sources. Despite the fact that they give a certain result, from the point of view of the code itself, they look a bit sloppy. So there was a desire when viewing listings to space, tabs and hyphenation. As they say, who clearly thinks - that clearly states. In this case, there was a feeling that the code itself is not completely understood by those who use it. I hope that this criticism will be perceived correctly and will be an incentive to improve the quality of writing code in the future.

2. Rewrote examples from the Internet.



It was much more pleasant to read them. They traced the fact that the author understands the algorithms of computer vision, built on the basis of OpenCV. They look like a retelling of what the author has already seen before. Nevertheless, the quality of the execution was not very different from what was in the works from the first group. Some of these examples were able to track only one object at a time.

3. And the third group is actually represented by just one job, which won first place in our competition.



In it, a man showed significant experience in using and developing computer vision algorithms. Where there is a clear understanding of where that line is between the qualitative result and the requirements for resources. If our software for analyzing a single video channel will require a couple of Xeon's, hardly anyone will want to use it.
This work went far beyond the existing examples. And what is most remarkable to her was a text document with arguments and explanations. The principle of operation at all stages of analysis. From subtracting the background to tracking objects.

The person who sent this work is called Oleg and we will be very happy if, as a result of our further cooperation, after returning from Sri Lanka, we get well-implemented video stream analysis modules for Ivideon Server. Let's turn the work sent to the contest into a reliable analytics subsystem of our application and we will develop it further.

It is too early to guess when we can present the finished analytics to our users. But we really hope that by the end of February we will offer most of the video analysis functions that are present in expensive professional systems. And we will do it at a good level.

Personally, I am very glad that, in general, all our expectations were met, we found a worthy candidate and Habr is still some cake. A resource that unites the thinking intellectual elite of our society. She who knows how to not only put pros and cons, but also think, reason, oppose and make decisions on her own. People who, in spite of everything, achieve excellent results each in their field. It's great that you are! Thank!
If you have ideas and suggestions on the organization of a new competition - we will be glad to see them in the comments!